Ep.228 Everybody Bought Copilot — Nobody Can Prove It Worked | Denis O’Shea on Copilot ROI, Security & Agents

Josh Lupresto (00:00)
Welcome to the podcast designed to fuel your success in selling technology solutions. I’m your host, Josh Lupresto SVP of Sales Engineering at Telarus and this is Next Level BizTech.

Everybody, welcome back. we’ve been talking a lot about AI. We’ve been talking robot dogs, we’ve been talking you know, foundational models, all kinds of crazy stuff. on today, we got a good friend of the show here, Denis O’Shea, founder and CEO of Mobile Mentor, on with us today to talk to about the title of today is Everybody Bought Copilot and Nobody Can Prove It Worked. Almost nobody. Denis, welcome on, man.

Denis O’Shea (00:41)
Thank you, Josh. Nice to be on your show. Thank you.

Josh Lupresto (00:46)
we got a lot of good stuff to cover today. so mobile mentor, right? Supplier in the Tolaris ecosystem. before we get to all of the cool things that you guys do, we got a lot to unpack there. Who is Denis? I want to hear your story. How did you get to where you’re at forming mobile mentor? All that good stuff. Start us out.

Denis O’Shea (01:06)
who is Denis? I’m an Irishman who has spent many years traveling the world. I live in seven different countries. so I’ve got a very unusual hybrid accent from that I’ve picked up over the years, I guess. But I’ve always been in technology, studied technology, worked in technology all my life. I’m I’m a lifelong student, always trying to understand where technology is going and and the work we do as a service organization is helping our customers.

unpack the full potential of the technology they buy. So I run this business today. And by the way, we’re we’re going through a rebrand. So very soon you’ll know us as mentor, not mobile mentor. Drop in the word mobile. I’m just going to be mentor.

And so the work we

Josh Lupresto (01:46)
All right. Breaking news.

Denis O’Shea (01:48)
do is helping customers unpack all the goodies that they purchase within their technology stack. So we’re very rarely telling customers they need more technology. We’re usually telling them you got plenty. In many cases you got too much. Let’s

make sure you’re getting a lot of value out of it and doing all the things you could be doing, taking your business to the next level.

Josh Lupresto (02:10)
What was the what was the impetus for we’re gonna come back and unpack Mobile Mentor with what you just kinda started to talk about. But what was the impetus for in how did you get to this point of starting Mobile Mentor? Was there a single aha moment? Was it a bunch of things, weird scenarios? Is there what what’s the thought there?

Denis O’Shea (02:25)
Hmm. There

there was actually. there was a kick in the ass moment. I lost a major sale which burnt part of my brain, I think, and I’ve been trying to heal it for the last 22 years. So I was running Nokia’s operation in Central Europe and Switzerland, selling mobile technology, mobile networks at the time. And we were in a meeting with the CEO of one of the large European carriers, and he steered me down. He says,

Why would we sign this contract and buy more technology off you guys when our customers are not using the technology we bought off you over the last couple of years? And I had no answer. And I left that meeting with my tail between my legs and went back to my upland leadership in Finland and we couldn’t figure out the answer. And and I left Nokia to go and try and solve it. And the problem as I saw it at the time was.

People were racing out and spending a thousand euro buying their first smartphone and all they were doing was sending text messages and making phone calls. It was pretty dumb usage of a smart device. And so I figured somebody has to do something about this. And if it’s gonna be, it’s it’s it’s gonna start with me. And so I set up a business to help people get their first smartphone working properly, getting their email working, which was very hard twenty two years

ago.

And getting their calendar syncing and transferring all their contacts and getting their music going and getting all the things working before we even knew the word app. This was way before the app store

Josh Lupresto (03:58)
Yeah.

Denis O’Shea (03:59)
and all that. So it was quite a technical job. And and we did that and it worked really well. And we did expand to that too, and we rolled that service out to a million people worldwide. Brazil, China, Australia, Switzerland, Japan, UK, all those countries. And literally sat down with a million people, one one at a time.

And showed them how to get their smartphone fully configured to be a productive business tool. So that’s the problem I’ve been solving through our business. And that’s where we started. Now we’re doing it for AI, of course. People are buying, you know, whether it’s copilot and co-work and getting all their tokens and all the things and trying to figure out how do I use this technology to be more productive. Same question, different answer to 22 years ago. They were buying smartphones, now they’re buying AI. but but the core mission that for us was helping people.

Unlock the full potential of the technology they’re buying. Unlock the full potential of it, not just 30%.

Josh Lupresto (04:54)
I i I

I mean it’s a good I mean the DNA makes so much sense for what you’re doing right now. It’s just such a natural it’s it’s great, just another problem to solve, right? I mean, a as you think about that, as we start to unpack a little more about what you guys do, is it is it really that much different from the playbook you were running when you started? Like

Denis O’Shea (05:13)
Mm-hmm.

Josh Lupresto (05:16)
obviously AI is fun and exciting and this is life changing and all all that good stuff, but does the playbook change dramatically?

Denis O’Shea (05:24)
You know, that’s a really good lens to look at it. It i it it doesn’t, because back then our playbook from a go-to-market perspective was we formed a partnership with a carrier, so like China Unicom or Telecom Italia or Telstra in Australia, and every day they sent us a list of all the people who had just got their first BlackBerry or their first Nokia or their first Sony Ericsson the previous day and who signed up on a high end data plan. And then we would call those people and say, Hey, we’re

Mobile mentor, would you like one of our people to come out and spend an hour with you and get everything working? That was the proposition then. Now the proposition is Microsoft comes to us almost every day and says, hey, we’ve just done a big co-pilot deal with this organization over here. They need to get value from that investment. Can you help? And we’re like, ooh, yeah. So it’s a similar playbook, different technology, different motions, but the playbook is not that dissimilar.

Josh Lupresto (06:22)
I love it. all right. So let’s let’s get into this then. we’ve we’ve talked a lot lately on the podcast about you know, here’s what Anthropic is doing, here is what OpenAI has done, here’s this kind of race, it’s token maxing, all these different things happening, the cost, you know, people the the the the new shadow and kind of runaway IT. And you know, we did touch a little bit on recently on you know Microsoft’s

In the same game of, hey, this is really hard. We’ve got to put FDEs and you know, we’re we’re we’re investing into that. So I I would love to unpack Copilot a little more from your perspective. So let’s say there’s this let’s bust a myth, right? The myth is, or or just the customer complain, or the Microsoft telling you is that everybody says they rolled out copilot. They either added it on the license or it’s inclusive in the license, but

what’s the harder question, right? What what do you hit with? Is it the ROI? How do you start that when they say, Hey, we deployed it, what now?

Denis O’Shea (07:27)
I would ask I would start by asking how far they’ve gone in deploying it. Because many people started with the free version, Copilot Chat, which you know had some utility value. Copilot has a lot more. Cowork has a lot more again. And Scout will have even more again. So we’re seeing the technology advance quite rapidly. And so most organizations are not as far along as they might think. It’s still very it’s still early stage for most organizations. Most are still

doing pretty basic stuff with it. And in most organizations, it’s really only maybe ten, fifteen, twenty percent of people who are doing anything in earnest with AI, with two exceptions. One, if you’re a an organization that’s writing code, like if you’re building a SaaS product, they’re all over it. You know the AI is producing most of the code for people nowadays. Or if you’re producing copy, if you’re you know a press release company or a a media company or generating marketing content or something.

AI is doing most of the work. But for most organizations, AI is is still in its early innings, we’ll say. And what we see is most organizations have not yet laid down the five foundations for safe, secure AI adoption at scale. We’re still we’re still at the early stages of getting the foundations right. And unless those foundations are in place, AI cannot scale.

And at least our Lisa cannot scale securely and safely.

Josh Lupresto (09:00)
Can we can we unpack in in your mind what what are you i as you kind of start that conversation, I don’t know if it’s an ROI thing or you you’re talking foundations, can we unpack what are the foundations

Denis O’Shea (09:11)
Sto.

Josh Lupresto (09:11)
from your from your perspective when you get into these threads?

Denis O’Shea (09:14)
ROI is one of them, but we consider that to be the last one. The first one we believe is the use case definition, being really clear around who is getting AI and why, and doing the analysis up front to understand how people work today with the tools they’ve got today, documenting that. So like understanding a day in the life of the role, and then and then having a hypothesis to say with AI tools and workflows, that person will work differently. Better, faster, lower cost, whatever.

Having a clear view on and treating each use case as an experiment and documenting those, stack ranking them, so we can say these are the ones we’re going to do first, which means we’re going to suspend these other ones. We might come back to them later. That’s one. The second is the data security. And I think everyone everybody listening to this today, probably hyper tuned into this particular problem. But you know, data retention policies, data security, DLP policies and data sensitivity labeling.

And data access controls. Those are the things that most organizations do not yet have right. And so having visibility where employees might be uploading company information to some LLM tool that was not acquired, you know, not paying a license for it. It’s Shadow AI. But one of your smart employees needs to get something done. And so they’re uploading a company document into a public large language model. Now it’s gone. It’s out in a public large language model.

And and so putting the right framework in place for security, that that’s foundation two. The third is the training, an empowerment framework, not just one on one training, but training for groups and training specific to a use case. So if we build out a use case and we build agents for people and and and and enable them to equip work differently, changing behavior is incredibly hard. Incredibly hard. Like I feel like

Josh Lupresto (11:06)
Yeah, yeah.

Denis O’Shea (11:07)
One of the things we learned by sitting down with a million people between two thousand and four and two thousand and twelve was we learned what drives adoption and what are the barriers to adoption. And some of that is burned deep into our DNA, understanding technology, the challenges of technology adoption and how you roll out any new technology. Couldn’t it might be copilot today, but yeah, it can be any new technology. How you get people to change where they save documents, you know, saving things to OneDrive instead

of their

hard drive and changing how they save their how they manage passwords and all that. Any kind of technology change is hard, takes longer than we expect. And laying down a good framework for that by use case. That’s the third one. The fourth one is putting the right plumbing in place for agents so that agents have the right identity, access controls, all the right security parameters and can be managed through their life cycle. And then the fifth is the ROI.

Having a way of empirically measuring how people are working differently today to how they were working yesterday, and where they’re using Copilot and how they’re using it, or where they’re using co-work to do work with automation that they were doing manually in the past. And then being able to quantify that and say, we know what it’s costing us to do the co-work thing, because we can see the token usage, but how do we now equate that to a value for the task completed? So

So those are the five, what we call the five foundations the use cases, the data security, the

Josh Lupresto (12:38)
Yeah.

Denis O’Shea (12:38)
empowerment, the agent framework, and the ROM.

Josh Lupresto (12:42)
I like that. I I love a good foundation like that. I just think anytime we can give our advisors this framework or the power questions or just kind of that sequence, right? I I I think we’re seeing a lot of hey, we deployed it, but nobody’s really challenging themselves to say, you know, did it did it follow those guidelines, right? We might have had a hundred percent excitement, but we had a three percent uplift in productivity for a ninety percent uplift on cost. Like maybe let’s think about that.

Right. So I love the I love the framework thought

Denis O’Shea (13:11)
Yes, yes. Yeah.

Josh Lupresto (13:13)
there.

Denis O’Shea (13:14)
And by the way, this framework came from our own mistakes. So we made all the mistakes in the book here in Mobile Mentor. We rolled out AI early. I w you know, as a chief exec, I was the chief enthusiast and evangelist, and I was pushing this and I was evangelizing early on. And then we got ourselves in trouble because we realized that we didn’t have the foundations in place. And

When we did our first data scan, you’d be horrified, but we had thirty-three thousand sensitive data assets that were overshared across the organization. Let’s say it was a smaller number of the second scan and the third scan and the fourth scan, because we got our we got our act together. but we have we a lot of the the a lot of the things we’re doing now were a result from having made the mistakes ourselves internally. So, you know, I when I talk to clients, I say, make new mistakes. Please don’t make the mistakes we made. Make new ones, push

Josh Lupresto (14:04)
Yeah.

Denis O’Shea (14:04)
the boat out further.

Go faster, do new things, but you don’t need to repeat our mistakes. We can we can we can solve for those ones pretty easily. So then you can go forward to make new ones.

Josh Lupresto (14:14)
Beautiful. you guys have a there’s there’s a study out there. I think it says here, I’m gonna pull up the stats. It said that sixty-seven percent of employees work around existing company policy, and then the three sixty five license use maybe averages forty four percent, right? So there’s some disparateness there in the numbers. What what do you think the leaders are missing about their own people with numbers like that? What does that mean?

Denis O’Shea (14:45)
Take the first one. the sixty-seven percent is appalling and embarrassing for most of us who are responsible for for driving technology. And what it means is that our employees are encountering friction during onboarding or during usage, or they don’t find our tools easy enough to use. So they are motivated to go and find other ways of getting the job done. And we hire smart, resourceful people, and they will find smart and resourceful ways to get the job done, even if they don’t like.

the friction or the security or the controls in our own organization. So the way that sh plays out, the way that it’s played out in the past was people using Dropbox and Gmail when they wanted to share some big files. The way

Josh Lupresto (15:27)
Yeah.

Denis O’Shea (15:27)
it’s playing out today is people using, you know, BYO AI tools. You got a personal account, chat or Deep Seek or whatever the thing is, Open

Josh Lupresto (15:37)
Yeah.

Denis O’Shea (15:37)
Claw, and you’re running these tools to do work for you because you don’t like what your company gave you or your company didn’t give you

Adequate tooling to get the job done the way you’d like to get it done. And the research we did, we sliced and diced by generation. So it’s actually the data is very different by generation. So Gen Z, we’ve got to love them, they are the most likely to work around the policies of an organization and work around their tools and go and do their own thing and take matters into their own hands. And as we go up through the ages, the baby boomer generation.

is least likely to do that. And so we look at Gen Z, young millennials, old millennials, Gen X, which is my generation, and then the boomers. And so there’s there’s direct correlation there. So the young workers are the ones who are actually putting us at risk. But we as I said before, we have to love them because the young workers are also the ones who are embracing AI the fastest and doing the most with it. And the older

Josh Lupresto (16:39)
Yeah. What a double edge, yeah.

Denis O’Shea (16:40)
workers are struggling to do more than, you know,

glorified search. So the the young workers are the ones building the agents, building the flows, figuring out the integrations, all the smart automations. So so I like to say that young generation, we need to harness their capability in a safe way because they will make us millions by figuring out smart ways of using AI, but they will also get us breached and cost us millions if we’re not careful. So there’s a really interesting dichotomy there, Doug.

We we as the adults

Josh Lupresto (17:13)
Yeah.

Denis O’Shea (17:13)
and as the leaders have to manage this and get our heads around it.

Josh Lupresto (17:17)
Yeah, fair.

Denis O’Shea (17:20)
Now the second thing you raised, the forty-four percent,

Josh Lupresto (17:21)
yeah. Yeah.

Denis O’Shea (17:23)
and all our data from years of doing assessments, is that most organizations are using somewhere in the mid forties, at like forty four, forty five percent of the capability they purchase from Microsoft. And we come at this with a Microsoft

Josh Lupresto (17:35)
Mm-hmm.

Denis O’Shea (17:37)
lens and we go through 120 topics to understand how they’re working today and what tools they’re using, and they’re only using about forty four, forty five percent of what they’ve purchased.

And they’ve also got lots of other overlapping tools. They might have CrowdStrike or Okta, Duo, and a whole bunch of other things. And they’re probably using 44, 45% of those as well. And that kind of leads back to our open hypothesis. You know, most of the time we find people have too much technology, partially deployed, partially integrated, and they’re not getting all the automations and capabilities they could be getting. That’s the opportunity.

Josh Lupresto (18:12)
Yeah. Makes

sense.

okay, so let’s my favorite part, right? My nerd hat here. let’s start to unpack unpack this stack a little bit, maybe maybe two parts to this. So I think you if you turn copilot on, right, it surfaces everything that somebody can already access. This starts to think a little bit about of permissions, right? Maybe unpack the the Microsoft stack here about how you think of, okay, I want to turn copilot on and I wanna go.

You’ve got a lot of you know, you’ve got a lot of products here, right? A lot of Microsoft products that you’ve got familiarity around. So maybe just walk us through, you know, where does that hey, I can just turn it on and it goes, you know, I I I would think they’re gonna secure me, right? Is this a purview problem? Is this intra? What walk me through that.

Denis O’Shea (19:03)
From a technology perspective, kind of so the way I think about it is I I I think about the top of the pyramid. So we think of Micr the Microsoft stack as a pyramid, and at the very top we think we think of identity. So the user’s identity will govern everything. And if identity is well architected and clean and uncluttered, life will be easy for the end user for everything they’re doing. All their sign ins, all their authentication, the every zero trust interaction will be easy.

And same for IT. If if identity is clean, life will be easy for IT. Conversely, if you’ve got a messy, convoluted, complex, cluttered identity architecture, life’s gonna be really painful for everybody with everything. So if we start with that and think about a cloud native identity configuration, we’re using UNTRA ID in the cloud. We’re creating dynamic security groups, and let’s say you join an organization.

Josh will belong to specific groups, and those groups will have will have specific attributes and permissions and policies. And everything you do will follow the inherited policies of the group or groups you belong to. So all the data searches you do, all the copilot actions you try to do, all the sharing you try to do, all your actions will be governed by the attributes in the groups you belong to. That’s the modern way of doing it.

And and those those can be extended over time. The old way of doing this was we had AD on-prem and we had static groups. And it was really hard to make changes and you know it became complex. So it’s actually really easy if we have an uncluttered environment and it’s all, it’s all in the cloud, and everything flows downstream from there, and you belong to groups. And then as we create agents, the agents will inherit our permissions. So I I built an agent, for example.

That pulls some data out of our CRM, something else out of our finance system, smashes it together with something out of our PSA tool, and produces a report that can be shared with some people. All the permissions to make that happen will be inherited from the groups I belong to. So getting that architecture right sounds easy when I explain it like this. Getting that right is the key to making this scale safely and securely.

So we’re not creating static groups and static permissions and one off API calls. It’s all following a dynamic logic.

Josh Lupresto (21:36)
okay, so if I think, you know, yeah, I think everybody kind of grew up in this, it’s Azure AD, it’s a mess. I don’t know, the all these different the forest and all these things. So I’ve got, to your point, I love the agents inherit the identity of the user. That makes

Denis O’Shea (21:50)
Yeah.

Josh Lupresto (21:51)
things logical and dynamic, that’s but but safe and secure. So if I think about okay, in the Microsoft stack, it kind of does have a lot of the buttons that I need, right? I’ve got intra. I’ve got

purview for data security and all the rules and you things you talked about around that.

Denis O’Shea (22:05)
Yes. Super important. Yeah.

Josh Lupresto (22:08)
And then, you know, we’ve got Defender, right, to take care of my endpoints

Denis O’Shea (22:13)
Yeah.

Josh Lupresto (22:14)
and my devices and all of that. And speaking of devices, I guess you roll in tune into that. I

Denis O’Shea (22:19)
Yes.

Josh Lupresto (22:20)
I know that’s not everything. And and I’m sure this is probably going to feel like a a a a layup of a home run for you here, but a real honest question because it goes back to your data point earlier of if they’re using 40

you know, forty percent of

Denis O’Shea (22:34)
Yeah.

Josh Lupresto (22:35)
the Microsoft stack, then if I think about all the other things that people are using, JAMF, Workspace One, CrowdStrike, why don’t people what are they missing about going all in on the Microsoft stack? Is it they just don’t understand it, it’s nuanced and they need help rolling it out? And that’s obviously where you guys come in. Is it

They just grew up and maybe the Microsoft stack wasn’t there five years ago and now it is and they’ve been using these tools. Like what’s you see so much of this? What’s your thought here?

Denis O’Shea (23:05)
I would say that the last thing you mentioned is true in many cases. Microsoft tends to launch products early. And the product is usually immature and doesn’t do all the things that people would like it to do. So they try it, they get frustrated, and they go off and buy something that is more mature. And so you end up with a stack that does have JAMF to manage your Macs. And then years later, Microsoft matures Intune. So Intune can do Mac management almost as well as as JAMF.

And then you have exactly the same scenario play out for say MFA or EDR or CASB or any of the technology building blocks. Microsoft launches early, it’s an okay product, it takes them three to five years to make it a best in class top right quadrant product. And the way we think about it is, you know, sometimes it will make sense to have a non-Microsoft product if your strategy is best of breed.

And for the last 10 years, many CISOs in particular built a tech stack where they had the best of breed for each area, their CASB and their EDR and their all the things. And then there’s been a shift, and AI is accelerating the shift where people are going from best of breed to best of platform, recognizing that the integration between the tools and the automations between them and the intelligence is now more important than having the very best in every category. You might be better off having the second best.

But having the best set of integrations. So that as we think about agentic security, for example, that we have intelligence that can reason over all the tools and all the data and all the telemetry, that’s probably more valuable than having best tools in each area that don’t talk to each other. And security tools are by design not intended to leak information or you know, they don’t play nicely together, or if they do, it’s very superficial.

But you know, if you if you do decide you’re going to go all in with Microsoft, and let’s say if it’s a defender suite and you’ve got defender for endpoint, defender for cloud apps, defender for identity, defender for Office 365, then there’s there’s capability to reason over all of that in real time, detect something, block it, or remediate it before a human being was ever notified that that an issue happened. And then where this is going, which is super interesting, is

This idea of creating red, blue, and green agents to do that work for us. So the red agents deliberately trying to hack into the environment, the blue agents doing the deep investigation and threat hunting work, and the green agents doing the remediation work. And if everything is in the Microsoft estate, the agents can do their thing. But if a lot of your security is locked up on other vendors, then they’re stonewalled.

Josh Lupresto (25:54)
Interesting. good good lot of good tech stack there. It’s it’s it’s beautiful to see how long it’s you know, over these last couple of years, just what this thing has come to from a Microsoft perspective. So it’s good to see now.

Denis O’Shea (26:05)
It’s extraordinary. It’s extraordinary. And and the other

thing that’s that’s extraordinary is watching the volume grow. I saw I was watching a Microsoft webinar yesterday. They’re up to one hundred one hundred trillion telemetry signals per day is what they’re collecting

Josh Lupresto (26:19)
McCannish.

Denis O’Shea (26:20)
from all their customers worldwide. That number hurts my head. One hundred

Josh Lupresto (26:22)
She’s yeah.

Denis O’Shea (26:23)
trillion per day. From all

Josh Lupresto (26:25)
Gosh.

Denis O’Shea (26:26)
every login and every time we click on a link and every time we try and open a file and it’s kind of pressure

Josh Lupresto (26:30)
Yeah.

Denis O’Shea (26:30)
testing, do you have the rights to do this thing?

One hundred trillion. So it’s a good thing we have AI to reason over all that because how else would you make sense of it? Yeah.

Josh Lupresto (26:38)
Yeah, yeah. That’s a that’s a lot. No

kidding. okay, so i you know, it’s it’s been interesting to see how as everybody comes to market, you know, AWS was first to launch some of this, Microsoft launched some of this funding and and others, right? They and and even Google in this, they want to prove out

That these things are doable. So you you guys have a an a really good understanding, I think, ECIF, ECIF, if I remember right.

Denis O’Shea (27:08)
Mm-hmm. Mm-hmm.

Josh Lupresto (27:09)
Walk me through what does Microsoft want to do with those funds and why do the TAs care?

Denis O’Shea (27:16)
Microsoft’s agenda is very simple, it’s to drive adoption and consumption. So the last thing they want to do is sell a customer a license or you know sign on sign a customer on a contract and have the customer not use the products and not renew. That’s their worst case scenario. The best case scenario is the customer buys the products and gets them properly deployed and adopted and consumed and stays forever. So what Microsoft will do is they will invest

And give money to specific partners with specific skills to do the design and build and deployment of that technology to make sure the customer is sticky, that that technology, you know, is well adopted. And so that that bucket you’re referring to, that’s one of a few funding buckets and customer investment funds, E C I F, and that can be used to offset some of the initial deployment costs.

And so we use that extensively because of our status as a as a Microsoft partner. So let’s say somebody is doing a rollout. Let’s say if they’re going from Jamf to Intune, that’s a very common pathway these days. You’re going from CrowdStrike to Defender or whatever, or Copilot. Very easy to get ECIF for Copilot these days. So an organization wants to roll it out and lay down those five foundations we talked about before. we would submit an application to Microsoft. And most cases they’ll say, Yes, get on with it, and they’ll

Ask the customer sign off a POE, a proof of engagement, to prove we we did the work and they got a good outcome. And then we get measured on the back end to make sure that the customer actually had the things turned on and was all lit up correctly and they can see their telemetry that customers using this technology. So that so ESIF is one bucket and it’s a big bucket, and there’s a lot of governance and a lot of audits and a lot of scrutiny around it. There are other buckets that are easier to get access to MCI, Microsoft.

Commerce incentives, I think is what it means. And that’s a set amount of funding to drive a specific workload. So for purview, you know, a lot of customers lighting that up now to secure their data. Or in the past it was in tune, not so much anymore. Copilot, absolutely, and co-work, easy enough to get MCI funding. And it’s not large amounts, but it helps because we can say to the customer, you know, it’s gonna cost you X for us to do this piece of work. But we can, but if we do everything on

time.

And you complete the paperwork on time and all that, and we’ll get this much back from Microsoft so we can offset the cost to you. So it helps.

Josh Lupresto (29:46)
Yeah. I love that. I

I I love that because you know, one obviously the if if somebody’s using another technology, you know, they’re clearly incented to bring over, obviously.

Denis O’Shea (29:55)
Yes. Yes.

Josh Lupresto (29:58)
But where I think this helps is, you know, when we first started seeing A AI roll out a few years ago, I think the premise was, my gosh, this is so amazing. Everybody’s just gonna use it and it’s gonna change your life and they’re gonna buy a lot of it and it’s gonna be awesome.

Well, the reality then, if you look at the last couple of years, kind of the way the economy’s gone and just kind of weird ups and downs, the people that are trying to use these tools and are trying to adopt them adopt them now just maybe lost a couple of people who they used to lean on internally or they go, gosh, I really want to do this and I need to, you know, I need to follow Dennis’s five pillars and I need the ROI, but like how do I actually do that? So I love hearing programs like this. I really don’t think we take advantage of those anywhere near enough.

because it can it can offset the customer’s problem of I don’t

Denis O’Shea (30:46)
Yes.

Josh Lupresto (30:46)
have anybody I I want to do this. I just I can’t get there, right? And you’re going, here it is, here rock and roll, take it off our hands, go start paying us every

Denis O’Shea (30:52)
Here it is. Yeah. Yeah.

Josh Lupresto (30:54)
month. We did all the hard work for you. So I love that program, all those programs.

Denis O’Shea (30:58)
Yeah. Yeah. And look, Microsoft is incredibly generous with these programs. Honestly, it’s it’s it’s it’s insane. I was on that webinar I was on yesterday talking about some of the new funding programs and new funding buckets they’ve opened up and they’re very, very generous. Like there’s a lot of money out there to be

Josh Lupresto (31:15)
Love it.

Denis O’Shea (31:17)
taken and and used correctly, appropriately. I said that there’s a lot of scrutiny as well. So if you you know if you’re trying to take advantage of it and use it the wrong way.

Josh Lupresto (31:26)
And

Denis O’Shea (31:28)
That’s not gonna have a happy ending. But

Josh Lupresto (31:27)
Yeah. Yeah.

Denis O’Shea (31:30)
but if you know if you’re in the business f for the long haul, and we’ve been in the Microsoft ecosystem for over a decade, and if you play the game right and do the right things and and drive the right outcomes, then they are just they’re they’re so loyal to their to their loyal partners. Works both ways.

Josh Lupresto (31:48)
Love it. all right, as we get to maybe just final few thoughts here. let’s go through let’s go through an example. take me inside, real example, what did you find? And then, you know, after you solved the problems, of course, what were the problems? But really what did that customer maybe look like a year later, six months later?

Denis O’Shea (32:09)
specific to a co pilot example is what you’re looking for?

Josh Lupresto (32:12)
Any any I

any any of these kind of transformations that we’re talking about could be Copilot, could be Intune, could be just you you got a lot of products to kinda pick from. I mean just what what one shows kind of a good example of, hey, when we got involved, this is what we kind of saw differently after the fact that the customer never would have got there on their own.

Denis O’Shea (32:30)
Okay, I’ll give you two. one was a very large school district that had about a million devices, managed with four different tools. There was a bit of a bit of a cluster. and five and a half years later, we’ve got all the devices on Microsoft, Microsoft Intune, everything in Entra, everything on Defender, and we’ve managed to get Google, Apple, and Microsoft in the same room and playing nicely.

To help the school manage Chromebooks and Apple MacBooks and iPads and BYO little Google Netbooks and Windows, half a million Windows machines. I consider that a huge success for that school. The schools don’t have big budgets, but they gotta make things happen. And and that school had a really interesting problem with a huge group of homeless kids. Huge, ridiculously high number of homeless kids, which was

Which blew my mind. and these kids still need to turn up every day with a device that’s charged and ready to use. And so some of the management challenges there were just extraordinary. So that was that still that continues to be a an engagement and a customer I love working with. The more interesting and and topical one right now would be I’m thinking of a healthcare provider where they were just in love with Claude.

It was like claw this and claw that and look at what we built and look at how clever we are. And Microsoft was obviously keen to build on top of their E5 license and do Copilot and brought us in to help with some of the some of the rationalization and talk about the foundations and getting this right. And we we helped them to understand that Microsoft was probably going to win this game in the long term.

Even though they didn’t have the latest and greatest, shiniest new model. Clearly, you know, Claude is winning that that game

Josh Lupresto (34:29)
Yeah.

Denis O’Shea (34:29)
today, and models will come and go and all that. But Microsoft have the top and the bottom. And what I mean by that is the bottom is having at an AI tool that is grounded in every piece of work you’ve ever done, every email you’ve ever sent, every meeting you’ve ever attended and recorded, every file you’ve ever opened or touched, every team’s message, all that is in your tenant and associated with.

Josh Lepresto. That’s an immutable association with freaking terabytes of data, right? Everything you’ve ever touched. Copilot can see all that, reason over it, and write like you, act like you, and think like you, because it can see all that. So that’s the bottom. And the top, the reason why Microsoft will win out long term is the control plane. They have they’ve shown us they’ve got the control plane for identity with Antra, the control plane for all our devices, including BYO devices with Intune.

The control plane for all our security with Defender. It’s 29 Defender products now. The control plane now for data with purview and all our classification and labels and blah, blah, blah. Now they’ve got Agent 365, which is the control plane for all the agents and skills we’re going to build out. So the way I look at it is big pyramid. Microsoft owns the control plane in the enterprise, and they own the ground layer where all our all our stuff is, everything we’ve ever written or drafted or touched.

In between you’ll have models. And what Microsoft are doing, what we share with this client, is Copilot is simply an interface. Copilot is the UI to AI. The UI to AI. And you go into Copilot and you can select your model. I want Cloud, I want Opus, I want Sonnet, I want ChatGPT. You select the model for the task if you’re analyzing a spreadsheet or writing a document, whatever, doing some deep research. But copilot will simply be the interface to allow the best model.

To access your data and all your dirty laundry and all the things you’ve ever touched, but doing so in a safe way that follows all those security controls and groups and permissions and policies we talked about before and sit under that control plane. The models will be sandwiched in the middle between your dirty data and the control plane. And that’s why I think Microsoft will win this out in the same way they’ve won out many other tech battles over the last fifty two years or whatever.

Whatever time span has been. And so with that particular organization, the healthcare one, they were paranoid about security. They said to us, yes, to all the other things we were talking about, the five foundations, but they said security is priority one A. Nothing else matters until we get security sorted, because if we have one PHI leak, if we have one privacy breach,

Josh Lupresto (37:13)
Yeah.

Denis O’Shea (37:14)
if we have one thing go wrong,

Josh Lupresto (37:16)
It’s over.

Denis O’Shea (37:17)
it’s over, we’ll have to slam on the brakes.

And that stops the whole AI program for at least 12 months while we get our act together and deal with the lawyers. And so priority 1A. and so we had to go through all the painful work across, I think 11,000 SharePoint sites and you know, crazy amount of data. and and in and took SharePoint Advanced Management to its breaking points and then had to use purview and and and do all the things to get the data to a point where they were happy that now we

can start.

Push. We can start to roll AI out to specific groups, non-clinical first, and then clinical groups and start doing the things. so that’s an incomplete story because it’s still a work in progress, but it’s a good example of an organization who just has this frontal lobe burning priority to secure their data before allowing Gen Z to go nuts with the tools.

Josh Lupresto (38:14)
Yeah.

I you know, I I I love that example actually, and I love the thesis of how Microsoft may win out in this. And I don’t think that is getting the same level of publicity and news that the drama and the battles between open AI and Anthropic get. But when you really think about it like that, Microsoft being really smart, playing the long game, has done

Denis O’Shea (38:38)
yeah.

Josh Lupresto (38:38)
well in the desktop, has done well in licensing, and nothing has really disrupted that moat.

so it paints a really it paints a really optimistic picture for Microsoft when you frame it up that way. I love that framing.

Denis O’Shea (38:50)
I

i i thank you. And look, I’ve never said it out loud like the way I just explained it before with the dirty data on the bottom, the control bin at the top, but that’s how I’ve been thinking about it for a long time. And you said it well. They they won out on the desktop. Like Intune, when when we started working with Microsoft Intune eleven or twelve years ago, it was a dog of a product. Honestly,

Josh Lupresto (39:10)
Yeah.

Denis O’Shea (39:11)
it was so bad. All all my engineers were like wild when I signed the partnership with Microsoft

Josh Lupresto (39:16)
Yeah.

Denis O’Shea (39:16)
because it was so inferior.

to AirWatch and Mobile Iron and Citrix

Josh Lupresto (39:20)
Yeah.

Denis O’Shea (39:21)
at the time. but we had a vision that they would grow into the space and they have. And Gartner at the time predicted that Microsoft will will end up at 95% market share here because they’ll play the long game, they’ll maybe do one or two acquisitions. They didn’t. They’d built it all organically and they can now manage every endpoint. So for the first time in history, every endpoint, including virtual machines and BYO devices can be managed on one screen.

It’s taken them a decade, but they did it. They will

Josh Lupresto (39:51)
Yeah. Yeah.

Denis O’Shea (39:52)
do the same with AI. From the control plane to the dirty data on the bottom. They will own that and they will give us a secure, safe way of scaling AI inside our businesses.

Josh Lupresto (40:04)
Yeah. all right, final thoughts here. So let’s land this with we always gotta include our our guests’ favorite probing questions, right? To to to take for the advisors listening. So Denis, wrap us with the probing questions to uncover what the opportunities need to draw out right now. And then just kind of where do you think this goes in the next twelve months?

Denis O’Shea (40:35)
Probing questions will be around three things: data, agents, spend. There are three huge problems that most organizations do not have their heads around yet. Where the data goes and where the data comes from. So data security posture management, DSPM, data security posture management, that’s the next frontier for managed services in the security arena. The second is spend.

So the agent economy really took off on the first of July. I just saw the first internal report from Mobile Mentor last night on how much money we spent in the last month with agents. And I’ve had people telling me I’ve got this workflow working and it cost me $3.64 to produce this report. So we’re now starting to get our heads around what it costs us to burn these tokens to do a task. And organizations that have a bit of chess GPT and a bit of cloud and a bit of Copilot and a bit of this and a bit of that.

They’re about to be hit with this frickin’ tidal wave of tokens, credits, dollars, and different lic charging mechanisms. And some will be licenses and some will be API calls and access usage. And then if you’re building agents and in foundry, you’ll have other consumption charges. It’s about to get really messy really quickly, and nobody has a budget for this specifically or a way of managing it. So AI spin management.

Is going to become a big problem in the same way that mobile data management became a big problem about 15 or 18 years ago. But this is going to be way worse and way bigger. And the third one is going to be agent management. Because we’re starting to build the first agents and the first skills, which will behave like agents. And then soon we will have lots of agents. And then we’ll have agents building agents. And every agent is an independent software product with an identity.

Because it has permissions and it has access to specific applications and access controls. And it has a lifecycle because it will need to be patched and updated as all those access points and APIs change. And every one of them will have a life cycle in an organization. And the person who created it might leave. And then you have an orphaned agent that’s doing a piece of work. And so suddenly the challenge of managing agents will become enormous because agents will be the next mushroom cloud inside the enterprise.

And some will hopefully be built the right way, following good frameworks, some will not. And you’ll have rogue agents. And you know, bad things will happen if we don’t have the right plumbing in place from the get-go. So I would say to the TAs out there, ask your customers, you know, poke them right in the eye. Yeah, what are you doing about data security posture management with AI running over your data or reasoning over your data? What are you doing about anticipating and managing spend?

When you’ve got people consuming tokens and quacking APIs and using all these different products, how are you going to manage all that? And how are you going to produce reports that show what it’s actually costing us per person, per cost center, and per department? So those three layers per individual, per cost center, per department. And then the third one would be asking the customer, what are you doing to manage agents and the proliferation of agents and what those agents are doing to your data?

and how you’re gonna manage them as they grow. So data, agents, and spend.

Josh Lupresto (44:04)
I love it. I think it yeah.

I love it. It paints a good picture to your point exactly. Where is this going? agents agents creating agents. it’s a

Denis O’Shea (44:13)
yeah.

Josh Lupresto (44:13)
it’s a real problem and I’d love to say, this is so far away. This is like tomorrow. So it’s yeah.

Denis O’Shea (44:18)
No, it’s it’s it’s it’s it’s it’s yesterday, Josh, because

the data out of Anthropic is really interesting. eighty percent of the new code they’re producing is written by Claude itself. So AI is writing the next generation of AI. So it’s already happening.

Josh Lupresto (44:31)
Yeah. Yeah.

Denis O’Shea (44:33)
Their AI tool is writing their code. And the average coder is producing eight times more code this year than they were last year.

I know. So it’s already happening.

Josh Lupresto (44:40)
Crazy. Crazy. And n and nobody’s

reading it. Nobody I mean, very minimally, right? We’re just like, I it’s it’s Opus four point whatever, I just trust it. It’s great, right?

Denis O’Shea (44:50)
Yeah. Well what the great thing now is you can write the code, do your vibe coding, whatever, and then you can also use AI for testing your code. And then you can use AI for doing your documentation

Josh Lupresto (45:00)
Yeah.

Denis O’Shea (45:00)
and and r run your whole CI C D pipeline. so yeah.

Josh Lupresto (45:04)
No. I love it. Beautiful.

Beautiful, man. all right, Denis, I think that’s where we wrap it. Lots of good stuff. a a nice deep dive into Copilot, all the Microsoft pieces of it. and I think everybody should walk away with this going. Man, I can really understand how you guys can help somebody dive deep that does not quite have Microsoft mastered. So great stuff. I really appreciate you coming on.

Denis O’Shea (45:27)
Thank you. And look, it’s it’s a it’s a learning curve. We’re figuring this out every day. We’re building the plane in the air.

Josh Lupresto (45:35)
Why not? Building it on the ground is boring. So I’ll take you on that. I love it. all right, everybody. As always, don’t forget wherever you’re listening to Spotify, Apple, somewhere else, this drops every Wednesday. so make sure that you are subscribed, you get that, and do me a favor, go rate the podcast, give us some comments, that will, you know, let us know how we’re doing. And some of you have given some great feedback on the last couple episodes of things you want to see more of. So I believe that will wrap us up for today. I’m your host.

Josh Lupresto SVP of Sales Engineering at Telarus And this is Next Level BizTech. Everybody bought Copilot and nobody can prove it work. Denis O’Shea, founder and CEO of Mobile Mentire. Till next time.

Denis O’Shea (46:17)
Thank you, Josh.